User login
Development of a National Precision Oncology Program (NPOP) Dashboard Suite and Data Mart For Monitoring Somatic Molecular Testing Use
BACKGROUND
As of May 2023, the Veterans Affairs (VA) National Precision Oncology Program (NPOP) has provided somatic molecular testing for nearly 36,000 Veterans with cancer. Automated tools to monitor test use (locally and nationally) have only been available for NGS testing in advanced stage lung and prostate cancer. To better track utilization of NPOP supported testing across all cancer indications, and to support strategies to promote wider adoption, we developed an automated data architecture to monitor program operations. Here, we describe the development of the NPOP data mart and summarize the core components of the NPOP Somatic Molecular Testing dashboards.
METHODS
SQL Server Integration Services was used to build the backend data mart, which required the ingestion of vendor-specific XML data and subsequent harmonization with data found in the VA’s Corporate Data Warehouse (CDW). The NPOP somatic testing dashboards, developed using Power BI, are securely hosted in the cloud, and accessible through SharePoint.
DATA ANALYSIS
The NPOP dashboard suite displays key measures using descriptive statistics, including counts, proportions, means, medians, and standard deviations. To support the visualization of comparisons we leveraged stacked and clustered bar charts, and violin plots.
RESULTS
The NPOP data mart refreshes nightly providing users with near real-time data. The NPOP somatic testing dashboards include an all vendor combined report and sub-reports organized by vendors: Foundation Medicine, Personalis, and Personal Genome Diagnostics and Tempus. All reports contain four views with the ability to toggle between tests ordered or completed. For current NPOP vendors, patient level data and turnaround time views were developed. Data are stratified by test category (i.e., NGS Solid, NGS Liquid, Heme, IHC) and can be viewed longitudinally (i.e., across time) and filtered by test date, VA facility, patient demographics, and cancer characteristics (diagnosis, stage). To date, over 50,000 tests have been ordered (90% through Foundation Medicine).
IMPLICATIONS
The NPOP data mart and operational dashboards synthesizes vast amounts of data into a visually consumable format that supports monitoring the uptake and variation of somatic molecular testing services being offered across the VA.
BACKGROUND
As of May 2023, the Veterans Affairs (VA) National Precision Oncology Program (NPOP) has provided somatic molecular testing for nearly 36,000 Veterans with cancer. Automated tools to monitor test use (locally and nationally) have only been available for NGS testing in advanced stage lung and prostate cancer. To better track utilization of NPOP supported testing across all cancer indications, and to support strategies to promote wider adoption, we developed an automated data architecture to monitor program operations. Here, we describe the development of the NPOP data mart and summarize the core components of the NPOP Somatic Molecular Testing dashboards.
METHODS
SQL Server Integration Services was used to build the backend data mart, which required the ingestion of vendor-specific XML data and subsequent harmonization with data found in the VA’s Corporate Data Warehouse (CDW). The NPOP somatic testing dashboards, developed using Power BI, are securely hosted in the cloud, and accessible through SharePoint.
DATA ANALYSIS
The NPOP dashboard suite displays key measures using descriptive statistics, including counts, proportions, means, medians, and standard deviations. To support the visualization of comparisons we leveraged stacked and clustered bar charts, and violin plots.
RESULTS
The NPOP data mart refreshes nightly providing users with near real-time data. The NPOP somatic testing dashboards include an all vendor combined report and sub-reports organized by vendors: Foundation Medicine, Personalis, and Personal Genome Diagnostics and Tempus. All reports contain four views with the ability to toggle between tests ordered or completed. For current NPOP vendors, patient level data and turnaround time views were developed. Data are stratified by test category (i.e., NGS Solid, NGS Liquid, Heme, IHC) and can be viewed longitudinally (i.e., across time) and filtered by test date, VA facility, patient demographics, and cancer characteristics (diagnosis, stage). To date, over 50,000 tests have been ordered (90% through Foundation Medicine).
IMPLICATIONS
The NPOP data mart and operational dashboards synthesizes vast amounts of data into a visually consumable format that supports monitoring the uptake and variation of somatic molecular testing services being offered across the VA.
BACKGROUND
As of May 2023, the Veterans Affairs (VA) National Precision Oncology Program (NPOP) has provided somatic molecular testing for nearly 36,000 Veterans with cancer. Automated tools to monitor test use (locally and nationally) have only been available for NGS testing in advanced stage lung and prostate cancer. To better track utilization of NPOP supported testing across all cancer indications, and to support strategies to promote wider adoption, we developed an automated data architecture to monitor program operations. Here, we describe the development of the NPOP data mart and summarize the core components of the NPOP Somatic Molecular Testing dashboards.
METHODS
SQL Server Integration Services was used to build the backend data mart, which required the ingestion of vendor-specific XML data and subsequent harmonization with data found in the VA’s Corporate Data Warehouse (CDW). The NPOP somatic testing dashboards, developed using Power BI, are securely hosted in the cloud, and accessible through SharePoint.
DATA ANALYSIS
The NPOP dashboard suite displays key measures using descriptive statistics, including counts, proportions, means, medians, and standard deviations. To support the visualization of comparisons we leveraged stacked and clustered bar charts, and violin plots.
RESULTS
The NPOP data mart refreshes nightly providing users with near real-time data. The NPOP somatic testing dashboards include an all vendor combined report and sub-reports organized by vendors: Foundation Medicine, Personalis, and Personal Genome Diagnostics and Tempus. All reports contain four views with the ability to toggle between tests ordered or completed. For current NPOP vendors, patient level data and turnaround time views were developed. Data are stratified by test category (i.e., NGS Solid, NGS Liquid, Heme, IHC) and can be viewed longitudinally (i.e., across time) and filtered by test date, VA facility, patient demographics, and cancer characteristics (diagnosis, stage). To date, over 50,000 tests have been ordered (90% through Foundation Medicine).
IMPLICATIONS
The NPOP data mart and operational dashboards synthesizes vast amounts of data into a visually consumable format that supports monitoring the uptake and variation of somatic molecular testing services being offered across the VA.
Development of an Informatics Infrastructure and Frontend Dashboard for Monitoring Clinical Operations of the National TeleOncology Service
Background
Since inception, the Veterans Affairs (VA) National TeleOncology (NTO) service has monitored clinical operations through data tools produced by the Veterans Health Administration Support Service Center (VSSC). Unfortunately, pertinent data are spread across multiple reports, making it difficult to continually harmonize needed information. Further, the VSSC does not account for NTO’s hub and spoke clinical model, leading to inaccuracies when attempting to analyze unique encounters. To address these challenges, NTO partnered with the VA Salt Lake City Health Care System Informatics, Decision-Enhancement, and Analytic Sciences Center (IDEAS) to develop an informatics architecture and frontend NTO Clinical Operations Dashboard (NCOD). Here, we summarize our dashboard development process and the finalized key reporting components of the NCOD.
Methods
The VA Corporate Data Warehouse (CDW) serves as the primary data source for the NCOD. SQL Server Integration Services was used to build the backend data architecture. Data from the CDW were isolated into a staging data mart for reporting purposes using an extract, transform, load (ETL) approach. The frontend user interface was developed using Power BI. We used a participatory approach1 in determining reporting requirements. Stakeholders included the IDEAS dashboard development team and potential end users from NTO, including leadership, program managers, support assistants, and telehealth coordinators.
Results
The NCOD ETL is scheduled to refresh the data nightly to provide end users with a near real-time experience. The NCOD is comprised of the following four data views: clinic availability, team productivity, patient summary, and encounter summary. The clinic availability view summarizes clinic capacity, no shows, overbookings, and percent utilization. Relative value unit- based productivity is summarized in the team productivity view. The patient summary view presents aggregated data for veterans served by NTO, including geographic distribution, with patient-level drill down displaying demographics, cancer characteristics, and treatment history. Lastly, the encounter view displays utilization trends by modality, while accurately accounting for the hub and spoke clinical model.
Conclusions
An informatics architecture and frontend information display that is capable of synthesizing EHR data into a consumable format has been pivotal in obtaining accurate and timely insight into the demand and capacity of services provided by NTO.
- Esquer Rochin MA, Gutierrez-Garcia JO, Rosales JH, Rodriguez LF. Design and evaluation of a dashboard to support the comprehension of the progression.
Background
Since inception, the Veterans Affairs (VA) National TeleOncology (NTO) service has monitored clinical operations through data tools produced by the Veterans Health Administration Support Service Center (VSSC). Unfortunately, pertinent data are spread across multiple reports, making it difficult to continually harmonize needed information. Further, the VSSC does not account for NTO’s hub and spoke clinical model, leading to inaccuracies when attempting to analyze unique encounters. To address these challenges, NTO partnered with the VA Salt Lake City Health Care System Informatics, Decision-Enhancement, and Analytic Sciences Center (IDEAS) to develop an informatics architecture and frontend NTO Clinical Operations Dashboard (NCOD). Here, we summarize our dashboard development process and the finalized key reporting components of the NCOD.
Methods
The VA Corporate Data Warehouse (CDW) serves as the primary data source for the NCOD. SQL Server Integration Services was used to build the backend data architecture. Data from the CDW were isolated into a staging data mart for reporting purposes using an extract, transform, load (ETL) approach. The frontend user interface was developed using Power BI. We used a participatory approach1 in determining reporting requirements. Stakeholders included the IDEAS dashboard development team and potential end users from NTO, including leadership, program managers, support assistants, and telehealth coordinators.
Results
The NCOD ETL is scheduled to refresh the data nightly to provide end users with a near real-time experience. The NCOD is comprised of the following four data views: clinic availability, team productivity, patient summary, and encounter summary. The clinic availability view summarizes clinic capacity, no shows, overbookings, and percent utilization. Relative value unit- based productivity is summarized in the team productivity view. The patient summary view presents aggregated data for veterans served by NTO, including geographic distribution, with patient-level drill down displaying demographics, cancer characteristics, and treatment history. Lastly, the encounter view displays utilization trends by modality, while accurately accounting for the hub and spoke clinical model.
Conclusions
An informatics architecture and frontend information display that is capable of synthesizing EHR data into a consumable format has been pivotal in obtaining accurate and timely insight into the demand and capacity of services provided by NTO.
Background
Since inception, the Veterans Affairs (VA) National TeleOncology (NTO) service has monitored clinical operations through data tools produced by the Veterans Health Administration Support Service Center (VSSC). Unfortunately, pertinent data are spread across multiple reports, making it difficult to continually harmonize needed information. Further, the VSSC does not account for NTO’s hub and spoke clinical model, leading to inaccuracies when attempting to analyze unique encounters. To address these challenges, NTO partnered with the VA Salt Lake City Health Care System Informatics, Decision-Enhancement, and Analytic Sciences Center (IDEAS) to develop an informatics architecture and frontend NTO Clinical Operations Dashboard (NCOD). Here, we summarize our dashboard development process and the finalized key reporting components of the NCOD.
Methods
The VA Corporate Data Warehouse (CDW) serves as the primary data source for the NCOD. SQL Server Integration Services was used to build the backend data architecture. Data from the CDW were isolated into a staging data mart for reporting purposes using an extract, transform, load (ETL) approach. The frontend user interface was developed using Power BI. We used a participatory approach1 in determining reporting requirements. Stakeholders included the IDEAS dashboard development team and potential end users from NTO, including leadership, program managers, support assistants, and telehealth coordinators.
Results
The NCOD ETL is scheduled to refresh the data nightly to provide end users with a near real-time experience. The NCOD is comprised of the following four data views: clinic availability, team productivity, patient summary, and encounter summary. The clinic availability view summarizes clinic capacity, no shows, overbookings, and percent utilization. Relative value unit- based productivity is summarized in the team productivity view. The patient summary view presents aggregated data for veterans served by NTO, including geographic distribution, with patient-level drill down displaying demographics, cancer characteristics, and treatment history. Lastly, the encounter view displays utilization trends by modality, while accurately accounting for the hub and spoke clinical model.
Conclusions
An informatics architecture and frontend information display that is capable of synthesizing EHR data into a consumable format has been pivotal in obtaining accurate and timely insight into the demand and capacity of services provided by NTO.
- Esquer Rochin MA, Gutierrez-Garcia JO, Rosales JH, Rodriguez LF. Design and evaluation of a dashboard to support the comprehension of the progression.
- Esquer Rochin MA, Gutierrez-Garcia JO, Rosales JH, Rodriguez LF. Design and evaluation of a dashboard to support the comprehension of the progression.
Castration-Resistant Prostate Cancer—Not Only Challenging to Treat, but Difficult to Define
Purpose
Examine the impact of different definitions of castration resistance used to identify patients with castration-resistant prostate cancer (CRPC) using electronic health records (EHR).
Background
CRPC is a form of prostate cancer that is resistant to treatment with androgen deprivation therapy (ADT) and is associated with higher morbidity and mortality. Widely used guidelines like the Prostate Cancer Working Group 3 (PCWG 3), the American Urological Association (AUA), and many others differ in their definitions of castration-resistance. Until now, the feasibility of identifying CRPC using different definitions from EHR data has not been studied.
Methods/Data Analyisis
EHR data from the Veterans Health Administration (01/2006-12/2020) were used to identify veterans with CRPC according to the following criteria: 1) PCWG 3—a PSA increase ?25% from the nadir with a minimum rise of 2 ng/mL, while castrate (testosterone < 50 ng/mL); 2) AUA—2 consecutive PSA rises of ?0.2 ng/mL; 3) CRPC screening—a PSA rise of > 0.0 ng/mL within a window of 7–90 days.
Results
36,101 unique patients were identified using 1 of (or a combination of) the 3 CRPC criteria. Approximately 12,775 (35%) patients met all 3 criteria, while 8,589 (24%) were identified by AUA, 4,785 (13%) by CRPC screening, and 145 (0.4%) by PCWG3. A total of 8,377 (23%) patients met both the AUA and CRPC screening criteria, 1,219 (3%) patients met the AUA and PCWG3 criteria, and 211 (1%) met the PCWG3 and CRPC screening criteria.
Conculsions/Implications
Although several definitions can be used to identify CRPC patients, a combination of these definitions results in the greatest yield of CRPC patients identified using EHR data. Even though the PCWG3 criterion is frequently used in both clinical trials research and retrospective observational research, PCWG3 may miss many patients meeting other criteria and should not be used by itself when studying patients with CRPC identified from EHR data.
Purpose
Examine the impact of different definitions of castration resistance used to identify patients with castration-resistant prostate cancer (CRPC) using electronic health records (EHR).
Background
CRPC is a form of prostate cancer that is resistant to treatment with androgen deprivation therapy (ADT) and is associated with higher morbidity and mortality. Widely used guidelines like the Prostate Cancer Working Group 3 (PCWG 3), the American Urological Association (AUA), and many others differ in their definitions of castration-resistance. Until now, the feasibility of identifying CRPC using different definitions from EHR data has not been studied.
Methods/Data Analyisis
EHR data from the Veterans Health Administration (01/2006-12/2020) were used to identify veterans with CRPC according to the following criteria: 1) PCWG 3—a PSA increase ?25% from the nadir with a minimum rise of 2 ng/mL, while castrate (testosterone < 50 ng/mL); 2) AUA—2 consecutive PSA rises of ?0.2 ng/mL; 3) CRPC screening—a PSA rise of > 0.0 ng/mL within a window of 7–90 days.
Results
36,101 unique patients were identified using 1 of (or a combination of) the 3 CRPC criteria. Approximately 12,775 (35%) patients met all 3 criteria, while 8,589 (24%) were identified by AUA, 4,785 (13%) by CRPC screening, and 145 (0.4%) by PCWG3. A total of 8,377 (23%) patients met both the AUA and CRPC screening criteria, 1,219 (3%) patients met the AUA and PCWG3 criteria, and 211 (1%) met the PCWG3 and CRPC screening criteria.
Conculsions/Implications
Although several definitions can be used to identify CRPC patients, a combination of these definitions results in the greatest yield of CRPC patients identified using EHR data. Even though the PCWG3 criterion is frequently used in both clinical trials research and retrospective observational research, PCWG3 may miss many patients meeting other criteria and should not be used by itself when studying patients with CRPC identified from EHR data.
Purpose
Examine the impact of different definitions of castration resistance used to identify patients with castration-resistant prostate cancer (CRPC) using electronic health records (EHR).
Background
CRPC is a form of prostate cancer that is resistant to treatment with androgen deprivation therapy (ADT) and is associated with higher morbidity and mortality. Widely used guidelines like the Prostate Cancer Working Group 3 (PCWG 3), the American Urological Association (AUA), and many others differ in their definitions of castration-resistance. Until now, the feasibility of identifying CRPC using different definitions from EHR data has not been studied.
Methods/Data Analyisis
EHR data from the Veterans Health Administration (01/2006-12/2020) were used to identify veterans with CRPC according to the following criteria: 1) PCWG 3—a PSA increase ?25% from the nadir with a minimum rise of 2 ng/mL, while castrate (testosterone < 50 ng/mL); 2) AUA—2 consecutive PSA rises of ?0.2 ng/mL; 3) CRPC screening—a PSA rise of > 0.0 ng/mL within a window of 7–90 days.
Results
36,101 unique patients were identified using 1 of (or a combination of) the 3 CRPC criteria. Approximately 12,775 (35%) patients met all 3 criteria, while 8,589 (24%) were identified by AUA, 4,785 (13%) by CRPC screening, and 145 (0.4%) by PCWG3. A total of 8,377 (23%) patients met both the AUA and CRPC screening criteria, 1,219 (3%) patients met the AUA and PCWG3 criteria, and 211 (1%) met the PCWG3 and CRPC screening criteria.
Conculsions/Implications
Although several definitions can be used to identify CRPC patients, a combination of these definitions results in the greatest yield of CRPC patients identified using EHR data. Even though the PCWG3 criterion is frequently used in both clinical trials research and retrospective observational research, PCWG3 may miss many patients meeting other criteria and should not be used by itself when studying patients with CRPC identified from EHR data.